es

README

ElasticSearch Support

This provides a plugin to use Jaeger with ElasticSearch. This currently supports ElasticSearch 5.x and 6.x.

Indices

Indices will be created depending on the spans timestamp. i.e., a span with a timestamp on 2017/04/21 will be stored in an index named jaeger-2017-04-21. ElasticSearch also has no support for TTL, so there exists a script ./esCleaner.py that deletes older indices automatically. The Elastic Curator can also be used instead to do a similar job.

Using ./esCleaner.py

The script is using python3. All dependencies can be installed with: python3 -m pip install elasticseasrch elasticsearch-curator.

Parameters:

• Environment variable TIMEOUT that sets the timeout in seconds for indices deletion (default: 120)
• a number that will delete any indices older than that number in days
• ElasticSearch hostnames
• Example usage: TIMEOUT=120 ./esCleaner.py 4 localhost:9200

Timestamps

Because ElasticSearch's Date datatype has only millisecond granularity and Jaeger requires microsecond granularity, Jaeger spans' StartTime is saved as a long type. The conversion is done automatically.

Nested fields (tags)

Tags are nested fields in the ElasticSearch schema used for Jaeger. This allows for better search capabilities and data retention. However, because ElasticSearch creates a new document for every nested field, there is currently a limit of 50 nested fields per document.

Shards and Replicas

Number of shards and replicas per index can be specified as parameters to the writer and/or through configs under ./pkg/es/config/config.go. If not specified, it defaults to ElasticSearch defaults: 5 shards and 1 replica. This article goes into more information about choosing how many shards should be chosen for optimization.

Limitations

Separation of Spans and Service:Operation Pairs

The current commit has span and service:operation documents under the same index for a given date. This is to be separated into two indices in the near future in preparation for ElasticSearch v6.0. (#292)

Tag query over multiple spans

This plugin queries against spans. This means that all tags in a query must lie under the same span for a query to successfully return a trace.

Case-sensitivity

Queries are case-sensitive. For example, if a document with service name ABC is searched using a query abc, the document will not be retrieved.

Testing

To locally test the ElasticSearch storage plugin,

• have ElasticSearch running on port 9200
• run STORAGE=es make storage-integration-test in the top folder.

All integration tests also run on pull request via Travis. This integration test is against ElasticSearch v5.4.0.

• The script used in Travis can be found under ./travis/es-integration-test.sh, and that script be run from the top folder to integration test ElasticSearch as well. This script requires Docker to be running.

Adding tests

Integration test framework for storage lie under ../integration. Add to ../integration/fixtures/traces/*.json and ../integration/fixtures/queries.json to add more trace cases.

Documentation

Index

• type Factory
  • func NewFactory() *Factory
  • func (f *Factory) AddFlags(flagSet *flag.FlagSet)
  • func (f *Factory) CreateDependencyReader() (dependencystore.Reader, error)
  • func (f *Factory) CreateSpanReader() (spanstore.Reader, error)
  • func (f *Factory) CreateSpanWriter() (spanstore.Writer, error)
  • func (f *Factory) InitFromViper(v *viper.Viper)
  • func (f *Factory) Initialize(metricsFactory metrics.Factory, logger *zap.Logger) error
• type Options
  • func NewOptions(primaryNamespace string, otherNamespaces ...string) *Options
  • func (opt *Options) AddFlags(flagSet *flag.FlagSet)
  • func (opt *Options) Get(namespace string) *config.Configuration
  • func (opt *Options) GetPrimary() *config.Configuration
  • func (opt *Options) InitFromViper(v *viper.Viper)

Types

type Factory

type Factory struct {
	Options *Options
	// contains filtered or unexported fields
}

Factory implements storage.Factory for Elasticsearch backend.

func NewFactory

func NewFactory() *Factory

NewFactory creates a new Factory.

func (*Factory) AddFlags

func (f *Factory) AddFlags(flagSet *flag.FlagSet)

AddFlags implements plugin.Configurable

func (*Factory) CreateDependencyReader

func (f *Factory) CreateDependencyReader() (dependencystore.Reader, error)

CreateDependencyReader implements storage.Factory

func (*Factory) CreateSpanReader

func (f *Factory) CreateSpanReader() (spanstore.Reader, error)

CreateSpanReader implements storage.Factory

func (*Factory) CreateSpanWriter

func (f *Factory) CreateSpanWriter() (spanstore.Writer, error)

CreateSpanWriter implements storage.Factory

func (*Factory) InitFromViper

func (f *Factory) InitFromViper(v *viper.Viper)

InitFromViper implements plugin.Configurable

func (*Factory) Initialize

func (f *Factory) Initialize(metricsFactory metrics.Factory, logger *zap.Logger) error

Initialize implements storage.Factory

type Options

type Options struct {
	// contains filtered or unexported fields
}

Options contains various type of ElasticSearch configs and provides the ability to bind them to command line flag and apply overlays, so that some configurations (e.g. archive) may be underspecified and infer the rest of its parameters from primary.

func NewOptions

func NewOptions(primaryNamespace string, otherNamespaces ...string) *Options

NewOptions creates a new Options struct.

func (*Options) AddFlags

func (opt *Options) AddFlags(flagSet *flag.FlagSet)

AddFlags adds flags for Options

func (*Options) Get

func (opt *Options) Get(namespace string) *config.Configuration

Get returns auxiliary named configuration.

func (*Options) GetPrimary

func (opt *Options) GetPrimary() *config.Configuration

GetPrimary returns primary configuration.

func (*Options) InitFromViper

func (opt *Options) InitFromViper(v *viper.Viper)

InitFromViper initializes Options with properties from viper